The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill processes untrusted user input to formalize arguments and conduct research, creating a surface for indirect instructions.
Ingestion points: User-provided 'argument text' processed in SKILL.md.
Boundary markers: Absent; there are no specific delimiters or instructions to ignore embedded commands in the user input.
Capability inventory: The skill executes lean --stdin via subprocess and spawns research subagents with web access.
Sanitization: Absent; user content is interpolated directly into logic structures and research prompts.
Command Execution (LOW): The skill intentionally executes local commands (lean --version, lean --stdin) to verify logical proofs. While this is the primary purpose of the skill, executing user-derived logic through a compiler/prover is a known but restricted execution path.

argument-validator